investigating the role of dust in ice nucleation within clouds and … · 2020. 7. 31. · standing...

Atmos. Chem. Phys., 18, 8707–8725, 2018https://doi.org/10.5194/acp-18-8707-2018© Author(s) 2018. This work is distributed underthe Creative Commons Attribution 4.0 License.

Investigating the role of dust in ice nucleation within clouds andfurther effects on the regional weather system over East Asia –Part 1: model development and validationLin Su1 and Jimmy C. H. Fung2,3

1School of Science, Hong Kong University of Science and Technology, Hong Kong, China2Division of Environment, Hong Kong University of Science and Technology, Hong Kong, China3Department of Mathematics, Hong Kong University of Science and Technology, Hong Kong, China

Correspondence: Lin Su ([email protected])

Received: 12 August 2017 – Discussion started: 13 October 2017Revised: 9 May 2018 – Accepted: 5 June 2018 – Published: 21 June 2018

Abstract. The GOCART–Thompson microphysics schemecoupling the GOCART aerosol model and the aerosol-awareThompson–Eidhammer microphysics scheme has been im-plemented in the WRF-Chem to quantify and evaluate theeffect of dust on the ice nucleation process in the atmo-sphere by serving as ice nuclei (IN). The performance of theGOCART–Thompson microphysics scheme in simulatingthe effect of dust in atmospheric ice nucleation is then eval-uated over East Asia during spring, a typical dust-intensiveseason, in 2012. Based upon the dust emission reasonably re-produced by WRF-Chem, the effect of dust on atmosphericcloud ice water content is well reproduced. With abundantdust particles serving as IN, the simulated ice water mixingratio and ice crystal number concentration increases by 15and 7 % on average over the dust source region and down-wind areas during the investigated period. The comparisonwith the ice water path from satellite observations demon-strated that the simulation of the cloud ice profile is substan-tially improved by considering the indirect effect of dust par-ticles in the simulations. Additional sensitivity experimentsare carried out to optimize the parameters in the ice nucle-ation parameterization in the GOCART–Thompson micro-physics scheme. Results suggest that lowering the thresh-old relative humidity with respect to ice to 100 % for the icenucleation parameterization leads to further improvement incloud ice simulation.

1 Introduction

Dust aerosol is the second largest contributor to the globalaerosol burden (Textor et al., 2006), and it is estimated tocontribute around 20 % to annual global aerosol emissions(Tomasi et al., 2017). The Intergovernmental Panel on Cli-mate Change (IPCC) has recognized dust as a major compo-nent of atmospheric aerosols, which are an “essential climatevariable.” East Asia is a main contributor to the Earth’s dustemissions. It has been reported in previous studies that EastAsian dust contributes 25–50 % of global emissions, depend-ing on the climate of the particular year (Ginoux et al., 2001).

Dust in the atmosphere alters the Earth’s weather and cli-mate in certain ways. By reflecting, absorbing, and scatteringthe incoming solar radiation, dust can cause a warming ef-fect within the atmosphere and a cooling effect at the surfacelayer (Lacis, 1995), which is the direct effect of dust. Thesemi-direct effect of dust is related to the absorption of short-wave and longwave radiation by dust aerosol within clouds,leading to a warming of the surrounding environment andcausing a shrinking of cloud and a lower cloud albedo, thusmodifying the radiation budget (Perlwitz and Miller, 2010;Hansen et al., 1997). The dust–cloud interaction is also re-ferred to as the indirect effect of dust. Dust particles are rec-ognized as effective IN and play an important role in the icenucleation process in the atmosphere, directly affecting thedynamics in ice and mixed-phase clouds, such as the forma-tion and development of clouds and precipitation (Koehler etal., 2010; Twohy et al., 2009).

Published by Copernicus Publications on behalf of the European Geosciences Union.

8708 L. Su and J. C. H. Fung: Investigating the role of dust in ice nucleation within clouds – Part 1

To date, many studies have been conducted to evaluate thedirect radiative effect of dust aerosol using radiation schemesimplemented in numerical models all over the world (Malletet al., 2009; Nabat et al., 2015a; Ge et al., 2010; Hartmannet al., 2013; Huang et al., 2009; Bi et al., 2013; Y. Liu etal., 2011; J. Liu et al., 2011; Chen et al., 2017). Recently,the semi-direct effect of dust has been investigated in a fewstudies over different regions by applying various global andregional models (Tesfaye et al., 2015; Nabat et al., 2015b;Seigel et al., 2013). Unfortunately, due to the poor under-standing of the dust–cloud interactions in microphysics pro-cesses, quantifying the microphysical effect of dust remainsa difficult problem. Various ice nucleation parameterizationshave been implemented into global models to estimate theimportance of dust in atmospheric ice nucleation (Lohmannand Diehl, 2006; Karydis et al., 2011; Hoose et al., 2008;Zhang et al., 2014). However, most regional models are notcapable of estimating the indirect effect of dust, and veryrarely has work been done to assess the indirect effects ofdust on the weather system, especially over East Asia, whichis a major contributor to global dust emissions. Currently,only a few microphysics schemes considering the aerosol–cloud interaction are implemented in regional models. Inmost of these microphysics schemes only the cloud conden-sation nuclei (CCN) served by aerosols are considered (Perl-witz and Miller, 2010; Solomos et al., 2011; Miller et al.,2004), while IN are not treated or represented by a prescribedIN distribution (Chapman et al., 2009; Baró et al., 2015), andthe production of ice crystals is simplified by a function oftemperature or ice saturation. In reality, however, the num-ber of ice crystals that can form in the atmosphere is highlydependent on the number of particles that can act as IN, anddust is the most abundant aerosol that can effectively serve asIN and affect the formation and development of mixed-phaseand ice clouds in the atmosphere. This effect should not beneglected in numerical models, especially in the simulationsover arid regions during strong wind events (DeMott et al.,2003, 2015; Koehler et al., 2010; Lohmann and Diehl, 2006;Atkinson et al., 2013).

In 2014, the aerosol-aware Thompson–Eidhammer mi-crophysics scheme, which takes into account the aerosolsserving as CCN and IN, has been implemented into theWeather Research and Forecast (WRF) model and also theWeather Research and Forecast model coupled with Chem-istry (WRF-Chem), enabling the model to explicitly pre-dict the number concentration for cloud droplets and icecrystals (Thompson and Eidhammer, 2014). Therefore, theaerosol-aware Thompson–Eidhammer scheme is an ideal mi-crophysics scheme for evaluating the effect of dust in atmo-spheric ice nucleation processes. However, this scheme is notcoupled with any aerosol model in WRF-Chem. When theaerosol-aware Thompson–Eidhammer microphysics schemeis activated, the model reads in pre-given climatologicalaerosol data derived from the output of other global cli-mate models, which introduces large errors into the estima-

tion of the effects of dust in microphysical processes. Thisproblem can be solved by embedding a dust scheme intothe Thompson–Eidhammer scheme or by coupling the mi-crophysics scheme with WRF-Chem. Compared with WRF,WRF-Chem integrates various emission schemes and aerosolmechanisms for simulating the emission, transport, mixing,and chemical transformation of aerosols simultaneously withthe meteorology (Grell et al., 2013). Therefore, WRF-Chemis more capable of producing a realistic aerosol field by com-paring the performances of different emission schemes oraerosol mechanisms.

In light of the above, we aim to fully couple the aerosol-aware Thompson–Eidhammer microphysics scheme with theGoddard Chemistry Aerosol Radiation and Transport (GO-CART) model (Ginoux et al., 2001) in the WRF-Chem mod-eling system in this study, enabling WRF-Chem to simulta-neously simulate the effect of dust aerosol in ice nucleationprocesses during simulations. Based upon the implementa-tion, the performance of the coupled GOCART–Thompsonmicrophysics scheme in simulating the ice nucleation pro-cess involving dust particles was validated and the role thatEast Asian dust plays in the ice nucleation process in the at-mosphere was further investigated.

The remainder of the paper is presented as follows. Sec-tion 2 provides a description of the model, and the imple-mentation work for coupling the aerosol-aware Thompson–Eidhammer microphysics scheme and the GOCART aerosolmodel in WRF-Chem is elaborated in Sect. 3, followed by themodel configurations for numerical simulations in Sect. 4.Section 5 presents the observational data used to validatethe performance of the GOCART–Thompson microphysicsscheme. Section 6 presents the results and discussion, fol-lowed by the conclusions in Sect. 7.

2 Model description

WRF-Chem is an online coupled regional modeling system,which means that it can simultaneously simulate the meteo-rological field, the chemical field, and the interactions in be-tween (Grell et al., 2013). The chemical model contains sev-eral gas- and aerosol-phase chemical schemes. In this study,we focus on the GOCART model, a simple aerosol modelthat will be used for dust simulation.

2.1 GOCART aerosol model

GOCART is an aerosol model for simulating major tro-pospheric natural-source aerosol components, such as sul-fate, mineral dust, black carbon, organic carbon, and sea-saltaerosols (Ginoux et al., 2001; Chin et al., 2000). It has beenimplemented into WRF-Chem as a bulk aerosol scheme. GO-CART is a simple aerosol scheme that can predict the mass ofaerosol components, but does not account for complex chem-ical reactions. Therefore, it is numerically efficient in simu-

Atmos. Chem. Phys., 18, 8707–8725, 2018 www.atmos-chem-phys.net/18/8707/2018/

L. Su and J. C. H. Fung: Investigating the role of dust in ice nucleation within clouds – Part 1 8709

lating aerosol transport and thus applicable to cases withoutmany chemical processes, especially dust events. Typically,it requires 40 to 50 % more computational time by applyingWRF-Chem run with the GOCART aerosol model than thestandard WRF to produce the same period of simulation.

Shao’s dust emission scheme (Kang et al., 2011; Shao,2004, 2001; Shao et al., 2011) is one of the dust emis-sion schemes in the GOCART aerosol model and has beendemonstrated to exhibit superior performance in reproduc-ing the dust cycle over East Asia compared to other emissionschemes (Su and Fung, 2015). The Shao emission schemewas updated in WRF-Chem since version 3.8 released in2016 to produce five size bins for dust emission, with diame-ters of < 2, 2–3.6, 3.6–6.0, 6.0–12.0, and 12.0–20.0 µm andmean effective radii of 0.73, 1.4, 2.4, 4.5, and 8.0 µm.

2.2 Thompson–Eidhammer microphysics scheme

The Thompson microphysics scheme is a bulk two-momentaerosol-aware microphysics scheme that considers the mix-ing ratios and number concentrations for five water species:cloud water, cloud ice, rain, snow, and a hybrid graupel–hailcategory (Thompson et al., 2004). The updated Thompson–Eidhammer scheme is an aerosol-aware version of theThompson scheme (Thompson and Eidhammer, 2014),which incorporates the activation of aerosols serving as cloudcondensation nuclei and IN, and therefore it explicitly pre-dicts the number concentrations of CCN and IN, as well asthe number concentrations of cloud droplets and ice crys-tals. Hygroscopic aerosols that serve as cloud condensationnuclei are referred to as water-friendly aerosols, and thosenon-hygroscopic ice-nucleating aerosols are referred to asice-friendly aerosols. The cloud droplets nucleate from ex-plicit aerosol number concentrations using a look-up tablefor the activated fraction as determined by the predicted tem-perature, vertical velocity, number of available aerosols, andpredetermined values of the hygroscopicity parameter andaerosol mean radius.

In the Thompson–Eidhammer scheme, the ice nucleationprocess is triggered once the relative humidity with respectto ice (RHi) exceeds 105 %. Furthermore, when the relativehumidity with respect to water (RHw) is above 98.5 %, it iscounted as condensation and immersion freezing and cal-culated by DeMott’s parameterization scheme (DeMott etal., 2010); when RHw is below 98.5 %, it is treated as de-position nucleation and determined by the Phillips param-eterization scheme (Phillips et al., 2008). Both the DeMottscheme and the Phillips scheme are coupled with the con-centration of ice-friendly aerosols. In addition, the freezingof deliquesced aerosols using the hygroscopic aerosol con-centration is parameterized following Koop’s parameteriza-tion scheme (Koop et al., 2000), and it is coupled with theconcentration of water-friendly aerosols.

The DeMott parameterization scheme for determining thecondensation and immersion freezing in the Thompson–

Eidhammer microphysics scheme was proposed in 2010 (De-Mott et al., 2010, hereafter referred to as the DeMott2010scheme) based on combined data from field experiments at avariety of locations over 14 years. In the Demott2010 param-eterization, the relationship between the number concentra-tion of aerosol-friendly aerosols and ice-nucleating particles(INP) is as follows:

nIN,Tk = a(273.16− Tk)bn(c(273.16−Tk)+d)aero , (1)

where nIN,Tk is the INP number concentration at a tempera-ture of Tk , naero is the number concentration of ice-friendlyaerosols, and a, b, c, and d are constant coefficients equal to5.94×10−5, 3.33, 2.64×10−2, and 3.33×10−3, respectively.The parameterization was tested with various temperaturesand number concentrations of ice-friendly aerosols, yieldinga good performance in reproducing the ice crystal numberconcentration under conditions of a relatively low mixing ra-tio of water vapor or a low concentration of INP comparedwith field–experimental data. The relationship between thesimulated number concentrations of ice-friendly aerosols andINP is basically linear for concentrations of both of under1000 cm−3 (DeMott et al., 2010).

The above parameterization was further developed in 2015(DeMott et al., 2015, hereafter the DeMott2015 scheme) forconditions of a higher mixing ratio of water vapor or a higherconcentration of ice crystals based on the latest data fromfield and laboratory experiments. According to the updatedobservational data, INP concentration increases exponen-tially with the number concentration of ice-friendly aerosols,and existing aerosols with relatively low concentrations (lessthan 1000 cm−3) can produce a large number of INP (morethan 100 000 cm−3). The updated relationship between thenumber concentrations of ice-friendly aerosols and INP inthe DeMott2015 parameterization scheme is as follows:

nIN,Tk = cfnα(273.16−Tk)+βaero exp(γ (273.16− Tk)+ δ) , (2)

where α, β, γ , and δ are constant coefficients equal to 0,1.25, 0.46, and −11.6, respectively. The calibration factor cfranges from 1 to 6 and is recommended to be 3.

The number concentration of INP produced by the De-Mott2015 scheme is much higher than that produced by theDeMott2010 scheme, and the difference grows larger withdecreasing temperature and an increasing number concen-tration of ice-friendly aerosols (DeMott et al., 2015). Al-though the DeMott2015 scheme has been implemented inthe code of the Thompson–Eidhammer scheme, it cannot beused without modifying the code. Instead of using the De-Mott2010 scheme by default, we modified the code to call theDeMott2015 scheme in the Thompson–Eidhammer schemefor the condensation and immersion freezing in our simula-tions investigate ice nucleation involving dust.

Originally, the calibration factor cf is set to be 3 and thethreshold temperature is set to be −20 ◦C. For the ice nu-cleation process in the Thompson–Eidhammer scheme, the

www.atmos-chem-phys.net/18/8707/2018/ Atmos. Chem. Phys., 18, 8707–8725, 2018


number concentrations of both water-friendly aerosols andice-friendly aerosols are pre-given in the initialization of thesimulations and are derived from the climatological data pro-duced by global model simulations in which particles andtheir precursors are emitted by natural and anthropogenicsources and explicitly modeled with various size bins formultiple species of aerosols by the GOCART model. In theconsequent simulations, a fake aerosol emission is imple-mented by giving a variable lower boundary condition basedon the initial near-surface aerosol concentration and a sim-ple mean surface wind for calculating a constant aerosol fluxat the lowest level in the model. The number concentrationsof both water-friendly aerosols and ice-friendly aerosols arethen updated at every time step by summing up the fakeaerosol emission fluxes and tendencies induced by aerosol–cloud interactions. The limitation of the current aerosol-aware Thompson–Eidhammer scheme is that the aerosol pro-file generated from a fake emission cannot represent the re-alistic aerosol level all the time, especially over areas withcomplex weather such as East Asia, leading to errors in quan-tifying the indirect effects of aerosols.

Coupling the GOCART aerosol model with theThompson–Eidhammer microphysics scheme allowsthe model to explicitly evaluate the indirect effect of natural-source aerosols on the basis of a relatively realistic emissionproduction, for instance the effect of dust on ice nucleationduring severe dust episodes or a dust-intensive season.

3 Implementation of GOCART–Thompsonmicrophysics scheme

To investigate the real-time indirect effects of dust aerosolover East Asia, a new treatment was implemented intoWRF-Chem to couple the GOCART aerosol model and theThompson–Eidhammer microphysics scheme, namely theGOCART–Thompson microphysics scheme. To accomplishthis, WRF-Chem version 3.8.1 has been modified in the fol-lowing three steps.

3.1 Upgraded GOCART aerosol model

Currently, the GOCART aerosol model generates only themass concentration for aerosols but no number concentra-tions. However, the number concentrations of aerosols arerequired for a microphysics scheme to evaluate the indi-rect effects of aerosols. Therefore, modification was neededto provide information about the number concentrations ofaerosols from the mass concentration produced in the GO-CART aerosol model.

The aerosol mass concentration was converted into a num-ber concentration using the aerosol density and effective ra-dius for each size bin. Assuming that dust particles are spher-ical, the mass per dust particle (mp, µg−1) for a size bin canbe approximated through the mean effective radius (rdust, m)

and density (ρdust, kgm−3) for that size bin.

mp = ρdust×43×πr3

dust (3)

The number concentration of dust particlesN (kg−1) for sizebin n at a grid point (i,j,k) is then calculated by using thefollowing equation:

N(i,j,k,n)= C(i,j,k,n)/mp, (4)

where C(i,j,k,n) is the dust mass mixing ratio (µgkg−1)for size bin n at grid point (i,j,k). Summing up the aerosolnumber concentrations through all of the size bins gives atotal dust number concentration, which will be passed intothe Thompson–Eidhammer microphysics scheme. Note thatall of the dust particles are treated as ice-friendly aerosolsin this study and represented by a newly introduction vari-able, ice-friendly aerosol produced by the GOCART aerosolmodel (GNIFA).

GNIFA(i,j,k)=n∑i=1

N (i,j,k,n) (5)

3.2 GOCART–Thompson microphysics scheme

This part of the modification was to hook up the GOCARTaerosol model and the Thompson–Eidhammer microphysicsscheme.

Instead of reading in the pre-given climatological aerosoldata, the initialization module of the Thompson–Eidhammermicrophysics scheme was modified to apply the bulk num-ber concentration of ice-friendly aerosols produced by theGOCART aerosol model for the calculation of the numberconcentration of ice-nucleating particles.

After the microphysical processes are finished for a partic-ular time step, the tendency of the bulk aerosol number con-centration (tendust, kg−1 s−1) produced by the microphysicsscheme is then passed into a wet scavenging scheme, whichwill be described in detail in the following subsection, for themodel to calculate the loss of aerosol mass due to the micro-physical processes within clouds and update the aerosol massfield.

3.3 In-cloud wet scavenging

As no in-cloud scavenging is considered for dust aerosol inWRF-Chem, a new wet scavenging process was introducedinto WRF-Chem to calculate the loss of aerosol mass dueto the microphysical processes within clouds using the ten-dency of aerosol number concentration produced by the mi-crophysics scheme. Assuming that the collection of dust par-ticles is proportional to the number concentration of dust par-ticles, the fraction of dust particles for each size bin (ϕ, %)can be calculated in the GOCART aerosol model.

ϕ(i,j,k,n)=N(i,j,k,n)

GNIFA(i, j, k)(6)



Figure 1. Nested domain set for the simulations. Blue dots repre-sent the 10 monitoring stations used for model validation. TD: theTaklimakan Desert; GD: the Gobi Desert.

The tendency of ice-friendly aerosol is then distributed intoeach size bin and the loss of dust mass due to the microphys-ical processes (wetscav, µgkg−1) for a particular size bin n iscalculated by using the following equation:

wetscav(i,j,k,n)= tendust(i,j,k)×ϕ(i,j,k,n)×mp× dt,(7)

where dt is the time step for the simulation.The mass mixing ratio (C, µgkg−1) for dust aerosol in a

particular size bin n is then updated at the next time step.

Ct+1(i, j, k, n) = C

t(i, j, k, n)−wetscavt(i, j, k, n) (8)

Apart from the in-cloud scavenging, the below-cloud wet re-moval is calculated by the default wet deposition scheme inthe GOCART aerosol model, in which the wet removal ofdust is accomplished by a constant scavenging factor whenthere is precipitation (Duce et al., 1991; Hsu et al., 2009).

4 Model configurations

A numerical experiment was conducted to examine the per-formance of the newly implemented GOCART–Thompsonmicrophysics scheme in simulating the ice nucleation pro-cess induced by dust in the atmosphere. Two simulations

were carried out for the numerical test. One control run(CTRL) was simulated without dust and one test run (DUST)with dust. According to the observations, the dust events in2012 over East Asia were concentrated in mid-March to lateApril, and the satellite observations from mid-March to theend of April were available for model validation; therefore,the simulation period was from 9 March to 30 April 2012,with the first 8 days as “spin-up” time. Only the results from17 March to 30 April 2012 were used for the analysis. Thefinal reanalysis data provided by the United States NationalCenter of Environmental Prediction with a horizontal resolu-tion of 1◦ were used for generating the initial and boundaryconditions for the meteorological fields, and the simulationswere re-initialized every 4 days with the aerosol field beingrecycled, which means that the output of the aerosol fieldfrom the previous 4-day run was used as the initial aerosolstate for the subsequent 4-day run. The integration time stepfor the simulations was 90 s.

Two nested domains were used for the simulations, asshown in Fig. 1. The outer domain (domain 1) is in a hor-izontal resolution of 27 km and covers the entire East Asiaregion. The inner domain (domain 2) is in a horizontal res-olution of 9 km and covers the entirety of central to easternChina. Both domains have 40 vertical layers, with the toplayer at 50 hPa. The locations of the two major dust sources,the Taklimakan Desert (TD) and the Gobi Desert (GD), aremarked in Fig. 1.

In the GOCART–Thompson scheme, deposition nucle-ation is determined by the Phillips parameterization (Phillipset al., 2008), the freezing of deliquesced aerosols using thehygroscopic aerosol concentration is parameterized follow-ing Koop’s parameterization scheme (Koop et al., 2000), andthe condensation and immersion freezing is parameterized bythe DeMott2015 ice nucleation scheme. The new wet scav-enging scheme was used for the in-cloud wet scavengingof aerosols due to microphysical processes. The GOCARTaerosol model was applied to simulate aerosol processes (Gi-noux et al., 2001, 2004) and produce the number concentra-tion of dust particles in DUST. Shao’s dust emission (Kanget al., 2011; Shao et al., 2011) with soil data from the UnitedStates Geological Survey (USDA, 1993), which have beendemonstrated to have good performance in reproducing dustemissions over East Asia, were used to generate dust emis-sions in the simulations of TEST. The number concentrationof dust particles was then fed into the GOCART–Thompsonmicrophysics scheme and treated as ice-friendly aerosols forcalculating the condensation and immersion freezing involv-ing dust by the DeMott2015 parameterization scheme. In ad-dition, the pre-given climatological profiles applied in theoriginal Thompson–Eidhammer scheme (Thompson and Ei-dhammer, 2014) were used to provide the number concentra-tion of water-friendly aerosols for the freezing of deliquescedaerosols calculated by Koops’s parameterization scheme toconsider the background indirect effect of aerosols on ice nu-



cleation for the simulations of both CTRL and DUST in thisstudy.

Other important physical and chemical parameterizationsapplied for the simulations are as follows. The Mellor–Yamada–Janjic (MYJ) turbulent kinetic energy scheme wasused for the planetary boundary layer parameterization (Jan-jic, 2002, 1994); the moisture convective processes wereparameterized by the Grell–Freitas scheme (Grell and Fre-itas, 2014); the shortwave (SW) and longwave (LW) radia-tion budgets were calculated by the Rapid Radiative TransferModel for General Circulation (RRTMG) SW and LW radi-ation schemes (Mlawer et al., 1997; Iacono et al., 2008); thegravitational settling and surface deposition were combinedfor the aerosol dry deposition calculation (Wesely, 1989); asimple washout method was used for the below-cloud wetdeposition of aerosols (Duce et al., 1991; Hsu et al., 2006);and the aerosol optical properties were calculated based onthe volume-averaging method (Horvath, 1998).

5 Observations

5.1 Surface PM10 observations

Hourly observations of the surface concentration of partic-ulate matter with a diameter smaller than 10 µm (PM10) at10 environmental monitoring stations located in or surround-ing the dust source areas in East Asia were used to examinethe capability of the model to reproduce dust levels at theground surface during the simulation period. The 10 stations(indicated by blue dots in Fig. 1) were located in the follow-ing five cities: Jinchang (Gansu Province), Yinchuan (Qing-hai Province), Shizuishan (Ningxia Province), Baotou (InnerMongolia), and Yan’an (Shaanxi Province), with 2 stationsin each city.

5.2 AERONET AOD observations

The AERONET program is a ground-based aerosol remotesensing network for measuring aerosol optical properties atsites distributed around the globe. This program providesa long-term database of aerosol optical properties such asaerosol extinction coefficient, single-scattering albedo, andaerosol optical depth (AOD) measured at various wave-lengths. The observational data from two sites were availablefor comparison with the simulation results during the sim-ulation period in this study. One was Dalanzadgad, locatedto the north of the Gobi Desert in Mongolia, and the otherwas the Semi-Arid Climate and Environment Observatory ofLanzhou University (SACOL), located in Lanzhou, GansuProvince, China. The exact locations of the two AERONETsites are depicted by the red triangles in Fig. 1. All of themeasured data passed the quality control standard level 2with an uncertainty of ±0.01 (Holben et al., 2001).

5.3 Satellite data

5.3.1 Multi-angle Imaging SpectroRadiometer (MISR)

The MISR instrument aboard the Terra platform of theUnited State National Aeronautics and Space Administra-tion (NASA) has been monitoring aerosol properties globallysince 2000. It measures the aerosol properties in four narrowspectral bands centered at 443, 555, 670, and 865 nm, due towhich the aerosol properties even over highly bright surfaces,such as deserts, can be retrieved (Martonchik et al., 2004;Diner et al., 1998). In this study, the AOD data at 555 nmretrieved from the MISR level 3 products with a spatial reso-lution of 0.5◦ were used for comparison with the spatial dis-tribution of simulated AOD over East Asia during the inves-tigated period.

5.3.2 Moderate Resolution Imaging Spectroradiometer(MODIS)

The MODIS instruments aboard the Terra and Aqua plat-forms of NASA monitor Earth’s surface and provide globalhigh-resolution cloud and aerosol optical properties at a near-daily interval (Kaufman et al., 1997).

To retrieve aerosol information over bright surfaces, theDeep Blue algorithm was developed to employ retrievalsfrom the blue channels of the MODIS instruments, at whichwavelength the surface reflectance is very low such that thepresence of aerosol can be detected by increasing total re-flectance and enhanced spectral contrast (Hsu et al., 2006).By applying this algorithm, the AOD values at wavelengthsof 214, 470, 550, and 670 nm over bright surfaces can be re-trieved. In this study, the MODIS level 2 AOD data at 550 nmwith a spatial resolution of 10 km were used for comparisonwith the simulated AOD during the simulation period.

5.3.3 Cloud-Aerosol Lidar and Infrared PathfinderSatellite Observation (CALIPSO)

The Cloud-Aerosol Lidar and Infrared Pathfinder Satellite,which is aboard the Aqua platform of NASA, combines anactive light detection and ranging (lidar) instrument with pas-sive infrared and visible imagers to probe the vertical struc-ture and properties of thin clouds and aerosols around theglobe (Vaughan et al., 2004). It aims to fill existing gaps inthe ability to measure the global distribution of aerosols andcloud properties and provides three-dimensional perspectivesof how clouds and aerosols form, evolve, and affect weatherand climate. It measures high-resolution vertical profiles ofaerosol and cloud extinction coefficients globally at wave-lengths of 532 and 1064 nm. The atmospheric IWC is de-rived from the observational cloud extinction coefficients at532 nm (Winker et al., 2009). In this study, the vertical pro-files of CALIPSO IWC with a horizontal resolution of 5 kmand a vertical resolution of 60 m were applied to verify the



Figure 2. Time series of spatially averaged daily dust mass load (a)and daily number density of dust particles (b) over East Asia (do-main 1) during the simulation period.

performance of the model in simulating the vertical distribu-tion of atmospheric IWC.

6 Results and model validation

6.1 Dust over East Asia

The time series of daily average dust load over the entire EastAsia region (domain 1) during the simulation period is shownin Fig. 2a. In total four dust events occurred during the sim-ulation period, lasting from 18 to 25 March, 30 March to7 April, 9 to 19 April, and 22 to 29 April 2012. The casefrom 22 to 29 April was the most significant one, with a dailydust load double that of the other cases. The fraction of dailydust load for each size bin is also shown in Fig. 2a. The dustparticles in the fourth and fifth bins with effective diametersranging from 6 to 20 µm account for the major part (around60 %) of the total mass of dust aerosols.

The number concentrations of dust particles over East Asiawere vertically integrated to obtain the number density ofdust particles. As shown in Fig. 2b, the time series of thedaily average number density of dust particles over East Asiaduring the simulation period shows a similar distribution as

that for dust load; the noteworthy distinction between the twotime series lies in the fraction of each size bin. The two sizebins with the smallest diameters (no larger than 3.6 µm) ac-count for over 80 % of the total number of dust particles, andthe particles with diameters smaller than 6 µm account forover 95 % of the total number of dust particles, indicatingthat the smallest dust particles are the main source of ice-friendly aerosol to serve as IN in the atmosphere.

6.1.1 Surface PM10 concentration

To evaluate the performance of WRF-Chem in reproducingdust emissions over East Asia, the simulated surface PM10concentrations were compared with the observations fromthe 10 environmental monitoring stations located near dustsources and downwind areas (described in Sect. 5.1). Thetime series of the observed and simulated surface PM10 con-centrations during the simulation period are shown in Fig. 3.Note that the simulated PM10 concentrations were extractedfrom the nearest grid point to the geographical coordinates ofthe stations. The stations in the same city were assigned intoone group such that there are five groups in Fig. 3. Overall,the model shows a good performance in simulating the dustcycle at different locations, with the evolution and magnitudeof the daily mean PM10 concentration well captured at mostof the stations. The model tends to produce lower surfacePM10 concentration than those observed, as no other emis-sions were considered in the simulations apart from dust.However, the dust events on 21 March and 26 April wereoverestimated by the model at both stations in Shizuishan(Fig. 3c and d) and Yinchuan (Fig. 3i and j).

The performance statistics were computed from the dailyaverage simulated PM10 concentration from DUST andthe corresponding observations, as shown in Table 1. Themodel tends to produce lower surface PM10 concentrationsthan those observed, as no other emissions were consid-ered in the simulations. The mean bias (MB) ranged from−108.73 to 72.46 µgm−3, with a mean over all the stationsof −18.84 µgm−3. The mean error (ME) ranged from 46.07to 155.83 µgm−3, with a mean over all of the stations of107.24 µgm−3. The root mean squared error (RMSE) rangedfrom 64.78 to 317.73 µgm−3, with a mean over all of thestations of 181.28 µgm−3. The relatively large values of theMB, ME, and RMSE are mainly attributed to the fact that noother aerosol emissions were considered in the simulationsother than dust, while the surface PM10 concentration at themonitoring stations is influenced by aerosols emitted fromother sources, such as anthropogenic emissions. The corre-lation coefficient (r) ranged from 0.59 to 0.87, with an aver-age for all of the stations of 0.70. The comparisons betweenthe observed and simulated surface PM10 concentration in-dicates that the model is capable of reproducing the surfacedust concentration reasonably during dust events over EastAsia.



Figure 3. Time series of hourly observed and simulated surface PM10 concentrations at various environmental monitoring stations; r repre-sents the correlation coefficient between simulation results and observations.

6.1.2 AOD time series

To examine the performance of the model in reproducing thecolumn sum of dust in the atmosphere, the simulated AODvalues were compared with observations measured at twoAERONET sites during the simulation period, as shown inFig. 4.

The site at Dalanzadgad (Fig. 4a) is located in Mongolia tothe north of the Gobi Desert. Overall, the evolution and mag-

nitude of the AOD time series at Dalanzadgad were reason-ably reproduced by the model during the simulation period,despite the fact that the simulated AOD was overestimated atthe end of March and in mid-April compared to the observedvalues.

SACOL (Fig. 4b) is a site located in Lanzhou, GansuProvince, which is a typical downwind area for dust in China.The model showed a good performance in reproducing the



Table 1. Performance statistics for the model in simulating surface PM10 concentrations at environmental monitoring stations during thesimulation period.

City Station no. MB (µgm−3) ME (µgm−3) RMSE (µgm−3) r

BaotouXCNAQ77 −36.18 80.43 94.88 0.59XCNAQ79 −10.05 75.83 106.58 0.62

ShizuishanXCNAQ346 72.46 121.18 317.73 0.79XCNAQ347 17.64 147.95 294.71 0.75

JinchangXCNAQ340 −108.73 109.09 128.56 0.77XCNAQ342 −18.65 46.07 64.78 0.70

Yan’anXCNAQ335 −38.93 99.05 149.44 0.68XCNAQ336 −60.15 124.74 166.89 0.60

YinchuanXCNAQ344 33.97 112.26 240.27 0.87CN_1487 −39.62 155.83 249.00 0.62

Average −18.84 107.24 181.28 0.70

MB: mean bias; ME: mean error; RMSE: root mean squared error; r: correlation coefficient.

Figure 4. Time series of daily mean observed and simulated aerosol optical depths at Dalanzadgad (a) and SACOL (b); r represents thecorrelation coefficient between simulation results and observations.

time series of AOD at SACOL during the entire simulationperiod, with the evolution and magnitude of AOD well cap-tured.

6.1.3 AOD spatial distribution

The spatial distribution of monthly mean simulated AOD isalso compared with observed values from MODIS and MISRproducts in Fig. 5. Note that the high AOD values observed innorthern, eastern, and southern China and part of SoutheastAsia are attributed to the abundant anthropogenic emissions,while those high values in the circle area are mostly due todust events. The region with high AOD values in the westernpart of the circled area is TD, and the region with relativelylower AOD in the eastern part of the circled area is GD. TheAOD observed by MODIS showed high values at the dustsource region in both March and April of 2012, as shown inFig. 5a and b. The mean observed AOD over GD was lowerthan that over TD in both March and April, and the mean

observed AOD was higher in April than in March over bothdust source areas. The spatial patterns of AOD observed byMISR are similar to MODIS, with comparable mean valuesover GD. However, the mean AOD values over TD observedby MISR are 36 and 40 % lower than those by MODIS inMarch and April, respectively (Fig. 5c and d).

The spatial patterns for the mean simulated AOD weresimilar to the observed values in both months but closer tothose from MODIS, as shown in Fig. 5e and f. The modelshows a good capability in capturing the spatial character-istics of the AOD over the dust source areas. For example,the mean observed AOD was higher in the southern part ofTD than that in the northern part in March and showed anincrease from March to April over GD, both of which werecaptured by the model. The values of the mean simulatedAOD over the Gobi Desert (0.33 for March and 0.39 forApril) are comparable to the observational values from bothMODIS (0.30 for March and 0.32 for April) and MISR (0.31



Figure 5. Spatial distributions of monthly mean AOD from MODIS observations (a, b), MISR observations (c, d), and simulation results (e,f) for March (left column) and April (right column) of 2012.

for March and 0.34 for April), but the mean simulated AODvalues over TD (0.54 for March and 0.64 for April) are be-tween the values of the MISR observations (0.72 for Marchand 0.88 for April) and the MODIS observations (0.46 forMarch and 0.53 for April).

In summary, it was demonstrated that the dust emissionssimulated by WRF-Chem are reliable for further analysis bythe comparison between the simulation results and the ob-servations for surface PM10 concentrations, as well as thetemporal and spatial distributions of AOD values.

6.2 Cloud ice over East Asia

Dust particles are effective IN and play an important rolein ice nucleation in the atmosphere under appropriate condi-tions. With the large number of IN created by dust particlesemitted into the atmosphere, an increase in the number of icecrystals is expected in the results from DUST compared withthose from CTRL, after taking into account the effects of dustparticles in the GOCART–Thompson microphysics scheme.Figure 6 shows the overall comparison between the numberof grid points of simulated cloud ice mixing ratio and icecrystal number concentration in corresponding value bins (at

all model grids at hourly intervals) from CTRL and DUSTduring the entire simulation period.

As expected, the model produces a higher cloud ice mixingratio (Fig. 6a) and ice crystal number concentration (Fig. 6b)in DUST. The simulated cloud ice mixing ratio produced inDUST is substantially higher than that produced in CTRLthroughout all value bins, especially in those bins with valueslower than 0.05 gkg−1. Similarly, the simulated ice crystalnumber concentration produced in DUST tends to be higherthan that in CTRL in all value bins. The substantial increasein the simulated cloud ice mixing ratio and ice crystal num-ber concentration indicates that the enhancement of the icenucleation process induced by dust is successfully repro-duced by the newly implemented GOCART–Thompson mi-crophysics scheme during the simulation period.

6.2.1 Spatial distribution of ice water path (IWP)

The spatial distributions of the simulated IWP and ice crys-tal number density from CTRL and DUST in Fig. 7 fur-ther demonstrate the enhancement in cloud ice due to dustover East Asia. The IWP produced by CTRL was relativelyhigh over western and eastern China, as well as at the south-ern boundary of the simulation domain, with values as high



Figure 6. 2-D histogram of simulated cloud ice mixing ratio (a) and cloud ice crystal number concentration (b) from CTRL and DUST. Thecolor scales indicate the number of grid points in specific value bins of ice mixing ratio and ice crystal number concentration.

Figure 7. Spatial distributions for the temporal mean simulated cloud ice water path (a–c) and ice crystal number density (d–f) from CTRL(left column), DUST (middle column), and the difference between CTRL and DUST (right column) over East Asia (domain 1) during thesimulation period.

as 20 gm−2 (Fig. 7a). After considering the effect of dustin the ice nucleation process, the IWP produced by DUSTincreased substantially over dust sources and downwind ar-eas (Fig. 7b and c), with values higher by 5–10 gm−2. Themean IWP averaged over the domain during the simulationperiod was 9.33 gm−2 for DUST and 7.95 gm−2 for CTRL.As shown in Fig. 7d–f, the spatial pattern for the enhance-ment of ice crystal number density over East Asia was simi-lar to that for the IWP. The mean ice crystal number densityaveraged over the domain during the simulation period was2.91× 108 m−2 for DUST and 2.76× 108 m−2 for CTRL.

The mean IWP and ice crystal number density were in-creased by 15 and 8 % over vast areas of East Asia upon con-sidering the effect of dust in the ice nucleation process in thesimulation, and such an effect can reach as far as the openocean of the Western Pacific (Fig. 7b and e), as the outbreakof a cold high system over northeast Asia can bring quantita-

tive dust aerosol down to the Western Pacific or even furtherduring the dust season.

6.2.2 IWC during dust events

The vertical profile of the simulated IWC was also comparedwith the observation from CALIPSO during dust events. Asmentioned in Sect. 5.1, a total of four dust events occurredduring the simulation period, lasting from 18 to 25 March,30 March to 7 April, 9 to 19 April, and 22 to 28 April2012. As shown in Figs. 8 and 9, the performance of themodel in simulating the vertical profile of IWC was evalu-ated by comparing the observations measured at 06:00 UTCon 21 March, 18:00 UTC on 1 April, 18:00 UTC on 9 April,and 05:00 UTC on 23 April 2012 with the simulated profilesat the same hour.

CALIPSO measures the global distribution of aerosol andcloud properties by lidar, which uses a laser to generate vis-



Figure 8. Spatial distribution for simulated dust load and satellite scanning track (a, b); the simulated vertical profile of ice-friendly aerosol(GNIFA) number concentration (c, d), with the orography represented by the shaded area; the CALIPSO vertical profile of IWC (e, f); andthe simulated vertical profile of IWC from CTRL (g, h) and DUST (i, j) for the case on 21 March (left column) and 1 April (right column)of 2012.

ible light with a wavelength of 1 µm or less to detect smallparticles or droplets in the atmosphere. Therefore, CALIPSOinstruments are more sensitive to tenuous ice clouds and liq-uid clouds composed of small particles or droplets, whichare invisible to instruments using signals of near-infrared orinfrared wavelength to detect clouds. Moreover, the lidar sig-nal is attenuated rapidly in optically dense clouds that the in-frared or near-infrared signals can easily penetrate (Winkeret al., 2010). As a result, the CALIPSO observations of IWCare mostly at the locations where the temperature is lowerthan−40 ◦C and the altitude is greater than 6 km poleward to12 km equatorward, mostly without precipitating ice. Giventhe above considerations, the simulated IWC profiles com-pared with the CALIPSO observations are referred to as onlycloud ice in this section.

The simulated dust load over East Asia at 06:00 UTC on21 March 2012 is shown in Fig. 8a, in which the dust cov-ered vast areas from western to eastern China between 35 and45◦ N, and the orbit of the satellite passed through the areawith a heavy dust load at around 100◦ E. Along the satelliteorbit, the abundant dust particles were transported to as highas 10 km aloft (Fig. 8c). At this time, a high concentration ofIWC was observed along the satellite orbit at an altitude ofaround 10 km between 30 and 45◦ N (Fig. 8e). The simula-tion result from CTRL (Fig. 8g) shows that the model pro-duces some ice cloud at altitudes of 9–10 km between 35 and45◦ N, but with much lower IWC compared to the observa-tions. Nevertheless, considering the effect of dust on the icenucleation process in DUST results in a much higher IWC ataltitudes of 9–10 km between 35 and 45◦ N (Fig. 8i), whichis much more consistent with the observations. The compar-



Figure 9. As Fig. 8 but for the cases on 9 April (left column) and 23 April (right column) of 2012.

ison between the simulation results from CTRL and DUSTindicates that the high IWC observed by the satellite between30 and 35◦ N might be unrelated to microphysical processes,but instead due to strong convective motions over southernChina.

On 1 April 2012, central to eastern China was covered bya thick dust plume, and the orbit of the satellite passed be-tween 25 and 43◦ N along 120◦ E at 18:00 UTC (Fig. 8b).Dust particles were distributed vertically from the surfaceto over 8 km along the satellite orbit (Fig. 8d). A band ofhigh IWC was observed by the satellite at altitudes of 5 to10 km between 33 and 44◦ N (Fig. 8f), which was highly un-derestimated in the results of the CTRL run without dust. Incontrast, the observed band of high IWC was reproduced bythe model in DUST with much more consistent magnitude(Fig. 8j).

At 18:00 UTC on 9 April 2012, the satellite was scanningthe dust source over GD, which was covered by a thick dust

plume between 35 and 45◦ N (Fig. 10a), with dust particleslifted up to 10 km above the surface (Fig. 9c). High concen-tration of IWC was observed by the satellite at altitude from5 to 11 km between 30 and 45◦ N (Fig. 9e). In this case, themodel reproduced the high concentration of IWC at the ob-served location in the results from both CTRL and DUST,although the IWC was significantly underestimated in the re-sults from CTRL (Fig. 9g), while it was better reproduced inthe results from DUST (Fig. 9j).

Similar to the previous cases, the satellite was scanningalong the east coast of China at 05:00 UTC on 23 April 2012,when a dust plume was arriving from the dust sources andaffecting areas between 35 and 45◦ N (Fig. 9b), and dust par-ticles were distributed vertically from the surface to 10 kmalong the scanning track of the satellite (Fig. 9d). Along theorbit of the satellite, a band of high IWC areas was observedat altitudes between 5 and 12 km from 30 to 45◦ N (Fig. 9f).In the results from CTRL, the model reproduced the high



Figure 10. Vertical profiles for the mean observed IWC from CALIPSO and the simulated IWC from CTRL and DUST for dust events on21 March and 1, 9, and 23 April 2012.

IWC values at the correct locations, but with substantiallylower values (Fig. 9h); however, upon taking into accountthe effect of dust in the GOCART–Thompson microphysicsscheme, the high IWC areas were well reproduced by themodel with much more consistent values (Fig. 9j).

By comparing the satellite-observed and simulated verti-cal profiles of IWC during the various dust events, it wasdemonstrated that the model reproduces the enhancement ofIWC clouds in the middle to upper troposphere by taking intoaccount the effect of dust in the ice nucleation process, whichsubstantially improves the simulation of cloud ice.

6.2.3 Mean vertical profiles of IWC

The mean profiles of the observed IWC, as well as the sim-ulated IWC from CTRL and DUST for the four dust eventsdiscussed in Sect. 6.2.2, are shown in Fig. 10. Note that the“mean profile” of IWC is the average over the available datapoints for the IWC along the orbit of the satellite between 30to 45◦ N for each of the dust events shown in Figs. 8 and 9.

Compared with the results from CTRL, the vertical profileof the simulated IWC was substantially improved in DUSTfor each dust event, with the enhancement of the ice nucle-ation process well captured by the GOCART–Thompson mi-crophysics scheme. However, there were still discrepanciesbetween observations and the simulation results from DUST,and the magnitudes of the vertical IWC produced by themodel were always lower than the observed values.

For the cases on 21 March and 1 April, the peaks of IWCwere observed at 9.5 and 8 km, respectively, whereas the sim-ulated peaks of IWC were located at 7 and 7.5 km, respec-tively, with lower peak values. The lower peak value for thecase on 21 March was due to the missing high IWC observedbetween 30 and 45◦N in the simulation results (Fig. 8e andi), while the lower peak value for the case on 1 April was due

to the underestimation of the IWC around 35◦ N (Fig. 8f andj). The locations of the peaks of simulated IWC for the caseson 9 and 23 April were more consistent with the observedpeaks, but still possessed lower values due to the missing orunderestimated high IWC with respect to the observations.

6.3 Sensitivity test and discussion

As discussed in Sect. 6.2.3, the simulation of cloud ice isgreatly improved by considering the enhancement of the icenucleation process induced by dust, which is well capturedby the GOCART–Thompson microphysics scheme. How-ever, the IWC is still underestimated by the model duringdust events. To determine the reason for this limitation, nu-merical experiments were performed to investigate the sensi-tivity of simulated IWC to the parameters of the ice nucle-ation parameterization in the GOCART–Thompson micro-physics scheme.

6.3.1 Calibration factor cf

The calibration factor cf is an empirical tuning coefficientderived from observational data from field and laboratory ex-periments. It ranges from 1 to 6 and is recommended to be3 (DeMott et al., 2015), which was applied in the previoussimulations. Three other experiments were conducted to in-vestigate the sensitivity of the simulated IWC to cf valuesranging from 3 to 6.

The mean profiles of IWC from simulation results werecompared with the CALIPSO observations for the dustevents discussed in Sect. 6.2.2 and 6.2.3, as shown in Fig. 11.For the cases on 21 March and 1 April, changing cf did notresult in an increase in IWC; instead, the simulated IWC re-mained consistent for cf values varying from 3 to 6.

For the case on 9 April, the simulated IWC increased be-tween 6 and 9 km and was higher and closer to the observed



Figure 11. Vertical profiles for the mean observed IWC from CALIPSO and the simulated IWC with various cf for the dust events on21 March and 1, 9, and 23 April 2012.

profile when cf was equal to 5 and 6 compared to the case inwhich cf was set to 3 and 4.

For the case on 23 April, two peaks were observed in theprofiles of simulated IWC located at 7 and 10 km. By in-creasing cf values from 3 to 6, the simulated IWC remainedunchanged for the peak at 10 km, but gradually increased forthe peak at 7 km. The peak of the simulated IWC at 7 kmshould correspond to the observed peak between 6 and 8 km,which was slightly overestimated by the model.

In summary, increasing the calibration factor cf from 3 to6 does not necessarily lead to a significant variation in thesimulated IWC during dust events, and the model achievesa relatively better performance in reproducing the profile ofIWC when the cf is set to 5.

As ice nucleation occurs only in a supersaturated atmo-sphere with respect to water vapor, the ice nucleation pro-cess would be terminated in the GOCART–Thompson micro-physics scheme when the environmental RHi is lower thanthe threshold RHi, which was set to 105 % for the simula-tions in this study. The consistency in the simulated IWCwith increasing cf for the cases in Fig. 11 indicates that inthese cases, the environmental RHi had already reached be-low 105 % when cf was set to 3, meaning that the water va-por available for freezing into ice crystals has been consumedwith cf equal to 3. Therefore, increasing cf could not lead toa further increase in simulated IWC. Given the above, lower-ing the threshold RHi might result in an enhancement of theice nucleation process as well as the simulated IWC, whichwill be discussed in the following section.

6.3.2 Threshold of relative humidity

In this study, the threshold relative humidity to trigger theice nucleation process in the simulation was originally set to105 %, which was selected for the central lamina condition in

the laboratory experiments to derive the DeMott2015 ice nu-cleation scheme (DeMott et al., 2015). However, as reportedin other studies, the number of ice-nucleating particles startsto rise when the relative humidity exceeds 100 % (DeMottet al., 2011). Therefore, a sensitivity experiment was carriedout to investigate the response of simulated IWC to a lowerthreshold relative humidity.

The mean profiles of IWC from the simulation results werecompared with the CALIPSO observations for the aforemen-tioned dust events, as shown in Fig. 12. With the thresh-old relative humidity lowered to 100 %, the simulated IWCshowed an increase throughout the vertical profile with themost significant increase at the peaks, suggesting more con-sistency with the observations for all of the dust events, ex-cept the one on 1 April. In the case on 1 April, the simulatedIWC increased at lower altitudes than the observed peak, butslightly decreased right at the peak with lowering the thresh-old relative humidity to 100 %. Overall, the simulation ofIWC during dust events was significantly improved by low-ering the threshold relative humidity from 105 to 100 %.

7 Conclusions

A new treatment, the GOCART–Thompson scheme, was im-plemented into WRF-Chem to couple the GOCART aerosolmodel to the aerosol-aware Thompson–Eidhammer micro-physics scheme. By applying this newly implemented mi-crophysics scheme, the effect of dust on the ice nucleationprocess by serving as IN in the atmosphere can be quanti-fied and evaluated. Numerical experiments, including a con-trol run without dust and a test run with dust, were thencarried out to evaluate the performance of the newly imple-mented GOCART–Thompson microphysics scheme in simu-lating the effect of dust on the content of cloud ice over East



Figure 12. Vertical profiles for the mean observational IWC from CALIPSO and the simulated IWC with threshold RH values of 105 and100 % for the dust events on 21 March and 1, 9, and 23 April 2012.

Asia during a typical dust-intensive period by comparing thesimulation results with various observations.

Based on the GOCART aerosol model the model repro-duced dust emissions reasonably well by capturing the evo-lution and magnitude of the surface PM10 concentration atthe locations of various environmental monitoring stationsand the AOD at two AERONET sites. The spatial patterns ofthe mean AOD over East Asia during the simulation periodwere also consistent with satellite observations.

The effect of dust on the ice nucleation process was thenquantified and evaluated in the GOCART–Thompson micro-physics scheme. Upon considering the effect of dust in thesimulation, the simulated ice water mixing ratio and ice crys-tal number concentration over East Asia were 15 and 7 %higher than those simulated without dust, with the most sig-nificant enhancements located over dust source regions anddownwind areas.

Comparison between the vertical profiles of the satellite-observed and simulated IWC during various dust events in-dicated that the enhancement of cloud ice induced by abun-dant dust particles serving as IN is well captured by theGOCART–Thompson microphysics scheme, with the resultsfrom the simulation with dust much more consistent with thesatellite observations, although the IWC is generally under-estimated by the model.

Sensitivity experiments revealed that the simulated IWCwas not very sensitive to the calibration factor defined inthe DeMott2015 ice nucleation scheme, but the model de-livered a slightly better performance in reproducing the IWCwhen the calibration factor was set to 5. However, the simu-lated IWC was sensitive to the threshold relative humidity totrigger the ice nucleation process in the model and was im-proved by lowering the threshold relative humidity from 105to 100 %.

Data availability. The WRF-Chem outputs and surface PM10observations are available from the authors upon request([email protected]). The MODIS level 2 data can be down-loaded freely from the NASA LAADS website (https://ladsweb.modaps.eosdis.nasa.gov/search/order/2/MOD06_L2--61, last ac-cess: 18 June 2018). The MISR level 3 data can be obtained fromthe NASA Atmospheric Science Data Center (https://eosweb.larc.nasa.gov/project/misr/misr_table, last access: 18 June 2018). TheAERONET data can be obtained from the AERONET programwebsite (https://aeronet.gsfc.nasa.gov/, last access: 18 June 2018).

Competing interests. The authors declare that they have no conflictof interest.

Acknowledgements. We would like to acknowledge the provisionof the MODIS and the MISR observations by the Ministry ofEnvironmental Protection Data Center, U.S. National Center forAtmospheric Research (NCAR) and the CALIPSO data by theU.S. National Aeronautics and Space Administration (NASA) DataCenter. We thank the principal investigators and their staff forestablishing and maintaining the two AERONET sites used in thisstudy. We appreciate the assistance of the Hong Kong Observatory(HKO), which provided the meteorological data. Lin Su would liketo thank Georg Grell, Stuart McKeen, and Ravan Ahmandov fromthe Earth System Research Laboratory, U.S. National Oceanic andAtmospheric Administration for insightful discussions. Other dataused this paper are properly cited and referred to in the referencelist. All data shown in the results are available upon request. Thiswork was supported by NSFC/RGC grant no. HKUST631/05,NSFC-FD grant U1033001, and RGC grant 16303416.

Edited by: Corinna HooseReviewed by: Gregory Thompson and two anonymous referees


https://ladsweb.modaps.eosdis.nasa.gov/search/order/2/MOD06_L2-61

https://ladsweb.modaps.eosdis.nasa.gov/search/order/2/MOD06_L2-61

https://eosweb.larc.nasa.gov/project/misr/misr_table

https://eosweb.larc.nasa.gov/project/misr/misr_table

https://aeronet.gsfc.nasa.gov/


References

Atkinson, J. D., Murray, B. J., Woodhouse, M. T., Whale, T. F.,Baustian, K. J., Carslaw, K. S., Dobbie, S., O’Sullivan, D., andMalkin, T. L.: The importance of feldspar for ice nucleation bymineral dust in mixed-phase clouds, Nature, 498, 355–398, 2013.

Baró, R., Jiménez-Guerrero, P., Balzarini, A., Curci, G., Forkel, R.,Grell, G., Hirtl, M., Honzak, L., Langer, M., and Pérez, J. L.:Sensitivity analysis of the microphysics scheme in WRF-Chemcontributions to AQMEII phase 2, Atmos. Environ., 115, 620–629, 2015.

Bi, J., Huang, J., Fu, Q., Ge, J., Shi, J., Zhou, T., and Zhang, W.:Field measurement of clear-sky solar irradiance in Badain JaranDesert of Northwestern China, J. Quant. Spectrosc. Ra., 122,194–207, 2013.

Chapman, E. G., Gustafson Jr., W. I., Easter, R. C., Barnard,J. C., Ghan, S. J., Pekour, M. S., and Fast, J. D.: Couplingaerosol-cloud-radiative processes in the WRF-Chem model: In-vestigating the radiative impact of elevated point sources, At-mos. Chem. Phys., 9, 945–964, https://doi.org/10.5194/acp-9-945-2009, 2009.

Chen, S., Huang, J., Kang, L., Wang, H., Ma, X., He, Y., Yuan, T.,Yang, B., Huang, Z., and Zhang, G.: Emission, transport, and ra-diative effects of mineral dust from the Taklimakan and Gobideserts: comparison of measurements and model results, At-mos. Chem. Phys., 17, 2401–2421, https://doi.org/10.5194/acp-17-2401-2017, 2017.

Chin, M., Rood, R. B., Lin, S.-J., Müller, J.-F., and Thompson,A. M.: Atmospheric sulfur cycle simulated in the global modelGOCART: Model description and global properties, J. Geophys.Res.-Atmos, 105, 24671–24687,2000.

DeMott, P. J., Sassen, K., Poellot, M. R., Baumgardner, D., Rogers,D. C., Brooks, S. D., Prenni, A. J., and Kreidenweis, S. M.:African dust aerosols as atmospheric ice nuclei, Geophys. Res.Lett., 30, https://doi.org/10.1029/2003GL017410, 2003.

DeMott, P. J., Prenni, A. J., Liu, X., Kreidenweis, S. M., Petters, M.D., Twohy, C. H., Richardson, M., Eidhammer, T., and Rogers,D.: Predicting global atmospheric ice nuclei distributions andtheir impacts on climate, P. Natl. Acad. Sci. USA, 107, 11217–11222, 2010.

DeMott, P. J., Möhler, O., Stetzer, O., Vali, G., Levin, Z., Petters,M. D., Murakami, M., Leisner, T., Bundke, U., and Klein, H.:Resurgence in ice nuclei measurement research, B. Am. Meteo-rol. Soc., 92, 1623–1635, 2011.

DeMott, P. J., Prenni, A. J., McMeeking, G. R., Sullivan, R. C.,Petters, M. D., Tobo, Y., Niemand, M., Möhler, O., Snider, J.R., Wang, Z., and Kreidenweis, S. M.: Integrating laboratory andfield data to quantify the immersion freezing ice nucleation activ-ity of mineral dust particles, Atmos. Chem. Phys., 15, 393–409,https://doi.org/10.5194/acp-15-393-2015, 2015.

Diner, D. J., Barge, L. M., Bruegge, C. J., Chrien, T. G., Conel, J.E., Eastwood, M. L., Garcia, J. D., Hernandez, M. A., Kurzweil,C. G., and Ledeboer, W. C.: The Airborne Multi-angle ImagingSpectroRadiometer (AirMISR): instrument description and firstresults, IEEE T. Geosci. Remote, 36, 1339–1349, 1998.

Duce, R., Liss, P., Merrill, J., Atlas, E., Buat-Menard, P., Hicks, B.,Miller, J., Prospero, J., Arimoto, R., and Church, T.: The atmo-spheric input of trace species to the world ocean, Global Bio-geochem. Cy., 5, 193–259, 1991.

Ge, J., Su, J., Ackerman, T., Fu, Q., Huang, J., and Shi,J.: Dust aerosol optical properties retrieval and radiativeforcing over northwestern China during the 2008 China-US joint field experiment, J. Geophys. Res.-Atmos., 115,https://doi.org/10.1029/2009JD013263, 2010.

Ginoux, P., Chin, M., Tegen, I., Prospero, J. M., Holben, B.,Dubovik, O., and Lin, S. J.: Sources and distributions of dustaerosols simulated with the GOCART model, J. Geophys. Res.-Atmos., 106, 20255–20273, 2001.

Ginoux, P., Prospero, J. M., Torres, O., and Chin, M.: Long-termsimulation of global dust distribution with the GOCART model:correlation with North Atlantic Oscillation, Environ. Modell.Softw., 19, 113–128, 2004.

Grell, G., Peckham, S., Fast, J., Singh, B., Easter, R., Gustafson,W., Rasch, P., Wolters, S., Barth, M., and Pfister, G.: WRF-ChemV3. 5: A summary of status and updates, EGU General AssemblyConference Abstracts, EGU General Assembly 2013, 7–12 April2013, Vienna, Austria, EGU2013-11332, 11332, 2013.

Grell, G. A. and Freitas, S. R.: A scale and aerosol awarestochastic convective parameterization for weather and airquality modeling, Atmos. Chem. Phys., 14, 5233–5250,https://doi.org/10.5194/acp-14-5233-2014, 2014.

Hansen, J., Sato, M., and Ruedy, R.: Radiative forcing and climateresponse, J. Geophys. Res.-Atmos., 102, 6831–6864, 1997.

Hartmann, D., Tank, A., and Rusticucci, M.: IPCC fifth assessmentreport, climate change 2013: The physical science basis, IPCCAR5, 31–39, 2013.

Holben, B., Tanre, D., Smirnov, A., Eck, T., Slutsker, I., Abuhas-san, N., Newcomb, W., Schafer, J., Chatenet, B., and Lavenu, F.:An emerging ground-based aerosol climatology: Aerosol opticaldepth from AERONET, J. Geophys. Res.-Atmos., 106, 12067–12097, 2001.

Hoose, C., Lohmann, U., Erdin, R., and Tegen, I.: The global in-fluence of dust mineralogical composition on heterogeneous icenucleation in mixed-phase clouds, Environ. Res. Lett., 3, 025003,https://doi.org/10.1088/1748-9326/3/2/025003, 2008.

Horvath, H.: Influence of atmospheric aerosols upon the global ra-diation balance, Atmospheric Particles, 5, 62–63, 1998.

Hsu, N. C., Tsay, S.-C., King, M. D., and Herman, J. R.: Deep blueretrievals of Asian aerosol properties during ACE-Asia, IEEE T.Geosci. Remote, 44, 3180–3195, 2006.

Hsu, S. C., Liu, S. C., Arimoto, R., Liu, T. H., Huang, Y. T., Tsai, F.,Lin, F. J., and Kao, S. J.: Dust deposition to the East China Seaand its biogeochemical implications, J. Geophys. Res.-Atmos.,114, https://doi.org/10.1029/2008JD011223, 2009.

Huang, J., Fu, Q., Su, J., Tang, Q., Minnis, P., Hu, Y., Yi, Y.,and Zhao, Q.: Taklimakan dust aerosol radiative heating derivedfrom CALIPSO observations using the Fu-Liou radiation modelwith CERES constraints, Atmos. Chem. Phys., 9, 4011–4021,https://doi.org/10.5194/acp-9-4011-2009, 2009.

Iacono, M. J., Delamere, J. S., Mlawer, E. J., Shephard, M.W., Clough, S. A., and Collins, W. D.: Radiative forc-ing by long-lived greenhouse gases: Calculations with theAER radiative transfer models, J. Geophys. Res.-Atmos., 113,https://doi.org/10.1029/2008JD009944, 2008.

Janjic, Z. I.: The step-mountain eta coordinate model: Further de-velopments of the convection, viscous sublayer, and turbulenceclosure schemes, Mon. Weather Rev., 122, 927–945, 1994.


https://doi.org/10.5194/acp-9-945-2009

https://doi.org/10.5194/acp-9-945-2009

https://doi.org/10.5194/acp-17-2401-2017

https://doi.org/10.5194/acp-17-2401-2017

https://doi.org/10.1029/2003GL017410

https://doi.org/10.5194/acp-15-393-2015

https://doi.org/10.1029/2009JD013263

https://doi.org/10.5194/acp-14-5233-2014

https://doi.org/10.1088/1748-9326/3/2/025003

https://doi.org/10.1029/2008JD011223

https://doi.org/10.5194/acp-9-4011-2009

https://doi.org/10.1029/2008JD009944


Janjic, Z. I.: Nonsingular implementation of the Mellor–Yamadalevel 2.5 scheme in the NCEP Meso model, NCEP office note,437, 61, 2002.

Kang, J. Y., Yoon, S. C., Shao, Y., and Kim, S. W.: Compar-ison of vertical dust flux by implementing three dust emis-sion schemes in WRF/Chem, J. Geophys. Res.-Atmos., 116,https://doi.org/10.1029/2010JD014649, 2011.

Karydis, V., Kumar, P., Barahona, D., Sokolik, I., and Nenes, A.:On the effect of dust particles on global cloud condensation nu-clei and cloud droplet number, J. Geophys. Res.-Atmos., 116,https://doi.org/10.1029/2011JD016283, 2011.

Kaufman, Y., Tanré, D., Remer, L. A., Vermote, E., Chu, A., andHolben, B.: Operational remote sensing of tropospheric aerosolover land from EOS moderate resolution imaging spectrora-diometer, J. Geophys. Res.-Atmos., 102, 17051–17067, 1997.

Koehler, K. A., Kreidenweis, S. M., DeMott, P. J., Petters, M. D.,Prenni, A. J., and Möhler, O.: Laboratory investigations of theimpact of mineral dust aerosol on cold cloud formation, Atmos.Chem. Phys., 10, 11955–11968, https://doi.org/10.5194/acp-10-11955-2010, 2010.

Koop, T., Luo, B., Tsias, A., and Peter, T.: Water activity as the de-terminant for homogeneous ice nucleation in aqueous solutions,Nature, 406, 611–614, 2000.

Lacis, A.: Climate forcing, climate sensitivity, and climate response:A radiative modeling perspective on atmospheric aerosols, DahlWs. Env., 11–42, 1995.

Liu, J., Zheng, Y., Li, Z., Flynn, C., Welton, E. J., and Cribb,M.: Transport, vertical structure and radiative properties of dustevents in southeast China determined from ground and space sen-sors, Atmos. Environ., 45, 6469–6480, 2011.

Liu, Y., Huang, J., Shi, G., Takamura, T., Khatri, P., Bi, J., Shi, J.,Wang, T., Wang, X., and Zhang, B.: Aerosol optical propertiesand radiative effect determined from sky-radiometer over LoessPlateau of Northwest China, Atmos. Chem. Phys., 11, 11455–11463, https://doi.org/10.5194/acp-11-11455-2011, 2011.

Lohmann, U. and Diehl, K.: Sensitivity studies of the importance ofdust ice nuclei for the indirect aerosol effect on stratiform mixed-phase clouds, J. Atmos. Sci., 63, 968–982, 2006.

Mallet, M., Tulet, P., Serça, D., Solmon, F., Dubovik, O., Pelon, J.,Pont, V., and Thouron, O.: Impact of dust aerosols on the radia-tive budget, surface heat fluxes, heating rate profiles and convec-tive activity over West Africa during March 2006, Atmos. Chem.Phys., 9, 7143–7160, https://doi.org/10.5194/acp-9-7143-2009,2009.

Martonchik, J., Diner, D., Kahn, R., Gaitley, B., and Hol-ben, B.: Comparison of MISR and AERONET aerosol op-tical depths over desert sites, Geophys. Res. Lett., 31,https://doi.org/10.1029/2004GL019807, 2004.

Miller, R., Tegen, I., and Perlwitz, J.: Surface radiative forcing bysoil dust aerosols and the hydrologic cycle, J. Geophys. Res.-Atmos., 109, https://doi.org/10.1029/2003JD004085, 2004.

Mlawer, E. J., Taubman, S. J., Brown, P. D., Iacono, M. J., andClough, S. A.: Radiative transfer for inhomogeneous atmo-spheres: RRTM, a validated correlated-k model for the longwave,J. Geophys. Res.-Atmos., 102, 16663–16682, 1997.

Nabat, P., Somot, S., Mallet, M., Michou, M., Sevault, F., Dri-ouech, F., Meloni, D., di Sarra, A., Di Biagio, C., Formenti, P.,Sicard, M., Léon, J.-F., and Bouin, M.-N.: Dust aerosol radiativeeffects during summer 2012 simulated with a coupled regional

aerosol–atmosphere–ocean model over the Mediterranean, At-mos. Chem. Phys., 15, 3303–3326, https://doi.org/10.5194/acp-15-3303-2015, 2015a.

Nabat, P., Somot, S., Mallet, M., Sevault, F., Chiacchio, M., andWild, M.: Direct and semi-direct aerosol radiative effect on theMediterranean climate variability using a coupled regional cli-mate system model, Clim. Dynam., 44, 1127–1155, 2015b.

Perlwitz, J. and Miller, R. L.: Cloud cover increase with in-creasing aerosol absorptivity: A counterexample to the conven-tional semidirect aerosol effect, J. Geophys. Res.-Atmos., 115,https://doi.org/10.1029/2009JD012637, 2010.

Phillips, V. T., DeMott, P. J., and Andronache, C.: An empiri-cal parameterization of heterogeneous ice nucleation for multi-ple chemical species of aerosol, J. Atmos. Sci., 65, 2757–2783,2008.

Seigel, R. B., van den Heever, S. C., and Saleeby, S. M.: Mineraldust indirect effects and cloud radiative feedbacks of a simulatedidealized nocturnal squall line, Atmos. Chem. Phys., 13, 4467–4485, https://doi.org/10.5194/acp-13-4467-2013, 2013.

Shao, Y.: A model for mineral dust emission, J. Geophys. Res.-Atmos., 106, 20239–20254, 2001.

Shao, Y.: Simplification of a dust emission scheme andcomparison with data, J. Geophys. Res.-Atmos., 109,https://doi.org/10.1029/2003JD004372, 2004.

Shao, Y., Ishizuka, M., Mikami, M., and Leys, J.: Pa-rameterization of size-resolved dust emission and valida-tion with measurements, J. Geophys. Res.-Atmos., 116,https://doi.org/10.1029/2010JD014527, 2011.

Solomos, S., Kallos, G., Kushta, J., Astitha, M., Tremback,C., Nenes, A., and Levin, Z.: An integrated modeling studyon the effects of mineral dust and sea salt particles onclouds and precipitation, Atmos. Chem. Phys., 11, 873–892,https://doi.org/10.5194/acp-11-873-2011, 2011.

Su, L. and Fung, J. C.: Sensitivities of WRF-Chem to dust emissionschemes and land surface properties in simulating dust cyclesduring springtime over East Asia, J. Geophys. Res.-Atmos., 120,https://doi.org/10.1002/2015JD023446, 2015.

Tesfaye, M., Tsidu, G. M., Botai, J., and Sivakumar, V.: Mineraldust aerosol distributions, its direct and semi-direct effects overSouth Africa based on regional climate model simulation, J. AridEnviron., 114, 22–40, 2015.

Textor, C., Schulz, M., Guibert, S., Kinne, S., Balkanski, Y., Bauer,S., Berntsen, T., Berglen, T., Boucher, O., Chin, M., Dentener, F.,Diehl, T., Easter, R., Feichter, H., Fillmore, D., Ghan, S., Ginoux,P., Gong, S., Grini, A., Hendricks, J., Horowitz, L., Huang, P.,Isaksen, I., Iversen, I., Kloster, S., Koch, D., Kirkevåg, A., Krist-jansson, J. E., Krol, M., Lauer, A., Lamarque, J. F., Liu, X., Mon-tanaro, V., Myhre, G., Penner, J., Pitari, G., Reddy, S., Seland, Ø.,Stier, P., Takemura, T., and Tie, X.: Analysis and quantificationof the diversities of aerosol life cycles within AeroCom, Atmos.Chem. Phys., 6, 1777–1813, https://doi.org/10.5194/acp-6-1777-2006, 2006.

Thompson, G. and Eidhammer, T.: A study of aerosol impacts onclouds and precipitation development in a large winter cyclone,J. Atmos. Sci., 71, 3636–3658, 2014.

Thompson, G., Rasmussen, R. M., and Manning, K.: Explicit fore-casts of winter precipitation using an improved bulk micro-physics scheme. Part I: Description and sensitivity analysis,Mon. Weather Rev., 132, 519–542, 2004.


https://doi.org/10.1029/2010JD014649

https://doi.org/10.1029/2011JD016283

https://doi.org/10.5194/acp-10-11955-2010

https://doi.org/10.5194/acp-10-11955-2010

https://doi.org/10.5194/acp-11-11455-2011

https://doi.org/10.5194/acp-9-7143-2009

https://doi.org/10.1029/2004GL019807

https://doi.org/10.1029/2003JD004085

https://doi.org/10.5194/acp-15-3303-2015

https://doi.org/10.5194/acp-15-3303-2015

https://doi.org/10.1029/2009JD012637

https://doi.org/10.5194/acp-13-4467-2013

https://doi.org/10.1029/2003JD004372

https://doi.org/10.1029/2010JD014527

https://doi.org/10.5194/acp-11-873-2011

https://doi.org/10.1002/2015JD023446

https://doi.org/10.5194/acp-6-1777-2006

https://doi.org/10.5194/acp-6-1777-2006


Tomasi, C., Fuzzi, S., and Kokhanovsky, A.: Atmospheric Aerosols:Life Cycles and Effects on Air Quality and Climate, John Wiley& Sons, Weinheim, Germany, 2017.

Twohy, C. H., Kreidenweis, S. M., Eidhammer, T., Browell, E.V., Heymsfield, A. J., Bansemer, A. R., Anderson, B. E., Chen,G., Ismail, S., and DeMott, P. J.: Saharan dust particles nucle-ate droplets in eastern Atlantic clouds, Geophys. Res. Lett., 36,https://doi.org/10.1029/2008GL035846, 2009.

USDA: Soil Survey Manual, Revised Edition, United States De-partment of Agriculture Handbook No. 18, US Department ofAgriculture, Washington, D.C., 1993.

Vaughan, M. A., Young, S. A., Winker, D. M., Powell, K. A., Omar,A. H., Liu, Z., Hu, Y., and Hostetler, C. A.: Fully automated anal-ysis of space-based lidar data: An overview of the CALIPSO re-trieval algorithms and data products, Remote Sens.-Basel, 16–30,2004.

Wesely, M.: Parameterization of surface resistances to gaseous drydeposition in regional-scale numerical models, Atmos. Environ.,23, 1293–1304, 1989.

Winker, D. M., Vaughan, M. A., Omar, A., Hu, Y., Powell, K.A., Liu, Z., Hunt, W. H., and Young, S. A.: Overview of theCALIPSO mission and CALIOP data processing algorithms, J.Atmos. Ocean. Technol., 26, 2310–2323, 2009.

Winker, D. M., Pelon, J., Coakley Jr., J., Ackerman, S., Charlson,R., Colarco, P., Flamant, P., Fu, Q., Hoff, R., and Kittaka, C.: TheCALIPSO mission: A global 3D view of aerosols and clouds, B.Am. Meteorol. Soc., 91, 1211–1229, 2010.

Zhang, C., Wang, M., Morrison, H., Somerville, R. C., Zhang, K.,Liu, X., and Li, J. L. F.: Investigating ice nucleation in cirrusclouds with an aerosol-enabled Multiscale Modeling Framework,J. Adv. Model Earth. Sy., 6, 998–1015, 2014.


https://doi.org/10.1029/2008GL035846

investigating the role of dust in ice nucleation within clouds and … · 2020. 7. 31. · standing...

Documents